Purpose: To evaluate the accuracy of ultrasonography (US) in the detection of Crohn disease in adults by systematically reviewing both cohort studies (those including patients whose clinical characteristics were consistent with those caused by an inflammatory bowel disease) and case-control studies (those in which patients with Crohn disease were compared with patients with other bowel diseases or healthy control subjects).
Materials and methods: The MEDLINE, EMBASE, and Cochrane Library databases were used to retrieve all the cross-sectional studies that assessed the diagnostic accuracy of US against that of one of several predefined reference standards (ie, radiologic, endoscopic, or histologic findings). The studies that fulfilled the inclusion criteria were identified, and their methodological quality was evaluated. Of the 2860 primary studies identified, two case-control and five cohort series fulfilled the inclusion criteria. Statistical analysis was performed by using the summary receiver operating characteristic (SROC) model.
Results: The ranges of US sensitivity and specificity for the diagnosis of Crohn disease reported for the included series were 75%-94% and 67%-100%, respectively; the heterogeneity of these values prevented the calculation of a cumulative value. The SROC curve revealed a clear cutoff effect that depended on the chosen bowel wall thickness threshold. Sensitivity and specificity of 88% and 93%, respectively, were achieved when a bowel wall thickness threshold greater than 3 mm was used, and sensitivity and specificity of 75% and 97%, respectively, were achieved when a threshold greater than 4 mm was used.
Conclusion: US examination seems appropriate for confirming or excluding Crohn disease as a diagnosis in a clinical context characterized by a pretest probability of Crohn disease that ranges from 12% to about 60%. In particular, for Crohn disease limited to the ileum, US may represent a valid alternative to the small-bowel series, while for colonic involvement US may be useful in ruling out the diagnosis.
Copyright RSNA, 2005